AI

In near-infrared optical breast lesion screening and diagnosis systems, high-speed four-dimensional scanners can dynamically acquire tens of thousands of lesion images within a five-minute period. Currently, manual computer annotation is required to generate standard samples from these scanned breast lesion images, a process that depends heavily on physicians with clinical expertise. On average, a single physician can annotate only approximately ten samples per working day. As a result, this process is time-consuming and labor-intensive, and the collected samples often suffer from low accuracy, large variability, and limited diagnostic reliability. Several AI-based annotation tools, such as QuPath, HALO AI™, and X-AnyLabeling, have been developed to assist this process. However, these tools are primarily manual or semi-automated and are unable to provide rapid and high-precision recognition. To address these limitations, this study proposes a new AI-based method for the rapid, accurate, and fully automated detection and diagnosis of breast lesions. The proposed approach complements existing AI-based annotation and diagnostic methods by enabling automated detection and classification of breast lesion samples. The proposed system employs a deep learning–based classification framework to construct a professional-level AI diagnostic model. The system automatically generates diagnostic outputs based on the annotation criteria used by professional physicians, including positive/negative classification and accuracy metrics. Compared with conventional manual diagnostic methods, the proposed approach provides faster and more reliable diagnostic estimates for new patients. These results demonstrate the potential of the proposed AI-based method to advance automated breast lesion screening and diagnosis and to contribute to future research and clinical applications in this field. Full article

Recent advances in computer vision have yielded models with strong performance on recognition benchmarks; however, significant gaps remain in comparison with human perception. One subtle ability is to judge whether an image looks like a given object without being an instance of that object. We study whether vision–language models such as CLIP capture this distinction. We curated a dataset named RoLA (Real or LookAlike) of real and look-alike exemplars (e.g., toys, statues, drawings, pareidolia) across multiple categories, and first evaluate a prompt-based baseline with paired “real”/“look-alike” prompts. We then estimate a direction in CLIP’s embedding space that moves representations between real and look-alike. Applying this direction to image and text embeddings improves discrimination in cross-modal retrieval on Conceptual 12M, and also enhances captions produced by a CLIP prefix captioner. Full article

Benchmark-driven evaluation helps distinguish between planning quality and interface reliability when large language models are utilized for embodied reasoning in simulation. Our submission to the Embodied Agent Interface Challenge (EAI) is evaluated across four stages of the pipeline. These being goal interpretation, subgoal decomposition, action sequencing, and transition modeling. The tasks run in the BEHAVIOR and VirtualHome simulators, which use constrained action vocabularies, fixed-object inventories and symbolic state representations within a standard evaluation protocol. Our system accesses the OpenAI API using GPT-4.1 for BEHAVIOR, GPT-4.1-mini for VirtualHome, and GPT-5-mini in later exploratory experiments across both environments. The schemas for each task determine how the outputs are structured, and outputs are regenerated when they do not follow the specification. On the final public leaderboard, our system ranked eighteenth overall with a score of 57.92, achieving 68.88 on BEHAVIOR and 46.96 on VirtualHome. In this paper, we describe our approach and discuss what these observations suggest about the strengths and limitations of current language models when used for embodied reasoning. Full article

The process of manually scoring drawings for the Goodenough–Harris Draw-a-Person (DAP) test is time-consuming and labor-intensive. It is also prone to inconsistencies due to subjective interpretation. Keeping these drawbacks in mind, this study aims to introduce a hybrid model of automated analysis and scoring of DAP test results using a combination of deep learning and rule-based reasoning. The proposed model has two different modules: one for predicting ten visual anatomical features of drawings using a convolutional neural network (CNN), and another set of six rules for representing geometric and spatial relationships. The output of the CNN is converted to binary using thresholding and then concatenated with the results of heuristic rules to obtain a final set of sixteen features. The proposed model was also evaluated using five-fold cross-validation methods and a separate hold-out test set containing 948 labeled drawings. The evaluation using the five-fold cross-validation approach shows that the proposed approach maintains consistent performance with high average F1-scores for all primary anatomical features above 0.90. On the other hand, the evaluation using the hold-out test set revealed that the proposed approach achieved a high macro-average accuracy of 91.78% for all sixteen features. This implies that the proposed approach has a high degree of generalization capability for the problem domain. The proposed approach achieves almost-perfect scores for structurally prominent anatomical features such as the head, limbs, trunk-related relationships, and all heuristic-based features. Nevertheless, the proposed approach performs poorly for less visually distinguishable anatomical features such as the ears (average F1-scores ≈ 0.09–0.12) and the neck (average F1-scores ≈ 0.75). The evaluation results show that the proposed approach is efficient in approximating expert-level scoring with a considerable reduction in human effort. Nevertheless, some limitations exist in the proposed approach. First, the proposed approach is less robust for subtle anatomical features. Second, the proposed approach relies on heuristic thresholds for feature extraction. Third, the proposed approach equally weighs all sixteen features; however, this may not exactly match the actual DAP scoring system. Full article

Nearly 2 million stillbirths occur globally each year. These outcomes are often driven by disparities in healthcare access, especially in low- and middle-income countries, where limited resources and shortages of trained medical personnel further increase preventable risks. Addressing these challenges requires not only strengthening healthcare systems but also enhancing intervention strategies. In this context, the development of decision-support systems becomes essential to dynamically identify at-risk pregnancies and improve fetal outcomes. Therefore, we propose AI-FRS (Artificial Intelligence–Fetal Risk Prediction System), a decision support tool for fetal risk prediction, designed to classify fetal conditions as healthy or at risk, using clinical data from Mexican obstetric patients. AI-FRS is built upon seven distinct machine learning models, systematically evaluated through 127 first-order ensemble combinations using hard voting. To further enhance predictive performance, we assessed 32,752 second-order ensembles, constructed by combining top-performing first-order ensembles across recall, precision, and F1-score metrics. The final selected model, called BSOEM, achieved a robust F1-score of 0.812, providing a more balanced and robust decision-making framework than individual models or simple ensembles. Additionally, we conducted an interpretability analysis to identify the clinical variables with the greatest contribution to model predictions, strengthening the system’s transparency and potential clinical trust. AI-FRS features a user-friendly interface specifically designed to facilitate adoption by healthcare professionals. This provides a fast and clinically applicable AI tool for intrapartum and peripartum risk detection in obstetrics, supporting clinical decision-making and improving fetal health outcomes. Full article

Most state-of-the-art machine learning techniques revolve around the optimization of loss functions, making the choice of an objective critical to model performance and reliability. Although recent reviews discuss loss functions in specific domains or in deep learning settings, there is still no single reference that presents widely used losses across major task families within a unified formal setting and with consistent optimization-relevant property annotations. In this survey, we compile and systematize the most widely adopted loss functions for regression, classification, generative modeling, ranking, energy-based modeling, and relational learning. Our selection procedure combines seeding from foundational textbooks and prior surveys with cross-checking of highly cited literature and common implementations in mainstream machine learning frameworks. We introduce 52 loss functions and organize them into an intuitive taxonomy, summarizing their theoretical motivation, key mathematical properties, and typical application contexts, with compact appendix tables for quick lookup. This survey is intended as a resource for undergraduate, graduate, and Ph.D. students, as well as researchers seeking a structured reference for selecting and comparing loss functions. Full article

Federated learning (FL) enables collaborative model training across distributed devices and robotic systems while preserving data privacy, making it well-suited for swarm robotics and edge-device-powered intelligence. However, FL remains vulnerable to adversarial behaviors such as data and model poisoning, particularly in real-world deployments where detection methods must operate under strict computational and communication constraints. This paper presents a practical, real-world federated learning framework that enhances robustness to adversarial agents in a ROS2-based edge-device swarm environment. The proposed system integrates the Federated Averaging (FedAvg) algorithm with a lightweight average cosine similarity-based filtering method to detect and suppress harmful model updates during aggregation. Unlike prior work that primarily evaluates poisoning defenses in simulated environments, this framework is implemented and evaluated on physical hardware, consisting of a laptop-based aggregator and multiple Raspberry Pi worker nodes. A convolutional neural network (CNN) based on the MobileNetV3-Small architecture is trained on the MNIST dataset, with one worker executing a sign-flipping model poisoning attack. Experimental results show that FedAvg alone fails to maintain meaningful model accuracy under adversarial conditions, resulting in near-random classification performance with a final global model accuracy of 11% and a loss of 2.3. In contrast, the integration of cosine similarity filtering demonstrates effective detection of sign-flipping model poisoning in the evaluated ROS2 swarm experiment, allowing the global model to maintain model accuracy of around 90% and loss around 0.37, which is close to baseline accuracy of 93% of the FedAvg algorithm only under no attack with a very minimal increase in loss, despite the presence of an attacker. The proposed method also maintains a false positive rate (FPR) of around 0.01 and a false negative rate (FNR) of around 0.10 of the global model in the presence of an attacker, which is a minimal difference from the baseline FedAvg-only results of around 0.008 for FPR and 0.07 for FNR. Additionally, the proposed method of FedAvg + cosine similarity filtering maintains computational statistics similar to baseline FedAvg with no attacker. Baseline results show an average runtime of about 34 min, while our proposed method shows an average runtime of about 35 min. Also, the average size of the global model being shared among workers remains consistent at around 7.15 megabytes, showing little to no increase in message payload sizes between baseline results and our proposed method. These results demonstrate that computationally lightweight cosine similarity-based detection methods can be effectively deployed in real-world, resource-constrained robotic swarm environments, providing a practical path toward improving robustness in real-world federated learning deployments beyond simulation-based evaluation. Full article

Accurate and early identification of ocular diseases is essential to prevent vision impairment and enable timely medical intervention. In routine clinical practice, ophthalmologists rely on a structured diagnostic workflow that incorporates multiple imaging modalities to manually assess and diagnose ocular diseases. However, interpreting each modality requires significant clinical experience and can be time-consuming. These limitations can be effectively addressed through the application of AI (Artificial intelligence)-driven multimodal fusion techniques. In this study, we conducted an empirical investigation to assess the impact of different fusion strategies—including early, intermediate, and late fusion—on diagnostic performance, training requirements, and interpretability. The proposed methodology was evaluated using three publicly available datasets: FFA-Fundus (Fundus fluorescein angiography), GAMMA (Glaucoma Analysis and Multi-Modal Assessment), and OLIVES (Ophthalmic Labels to Investigate Visual Eye Semantics). Experimental results demonstrate that multimodal feature fusion improves disease detection performance. Although fused models typically required an increase in training parameters compared to single-modality models, they provided interpretability on par with that of individual single-modal networks. However, inference time increased by approximately 50% for multimodal architectures. These findings underscore the value of integrating diverse ophthalmic imaging modalities to enhance diagnostic accuracy in automated disease detection systems. At the same time, the results highlight that unimodal models containing highly discriminative features can also perform competitively, particularly when a single modality is sufficient for disease identification. Multimodal fusion provides the greatest benefit in scenarios where complementary information across modalities contributes distinct and non-redundant features. Furthermore, fusing all available modalities may not be optimal due to increased computational cost and reduced inference efficiency; thus, selective modality integration and lightweight fusion strategies are essential to balance accuracy, interpretability, and efficiency in clinical deployment. Full article

Autonomous navigation is essential for orchard transportation robots to support automated operations and precision orchard management. However, in trellis orchards, dense vegetation and complex canopy structures often degrade the stability of GNSS-based navigation in in-row environments. To address this issue, this study proposes a GNSS–vision integrated navigation framework for orchard transportation robots. The performance of GNSS-based navigation in out-of-row environments and vision-based navigation in in-row environments was experimentally evaluated under representative orchard operating conditions. In out-of-row areas, the robot employs GNSS-based path planning and trajectory tracking to achieve reliable navigation in relatively open, lightly occluded environments. During in-row navigation, a deep learning-based real-time object detection approach is used to detect tree trunks and trellis supporting structures. By integrating corner-point selection with temporal RANSAC-based line fitting, a stable orchard row structure is constructed to generate robust navigation references. The visual perception module serves as the front-end sensing component of the navigation system and is designed to be independent of specific object detection architectures, allowing flexible integration with different real-time detection models. Field experiments were conducted under various orchard layouts and growth stages. The average lateral deviation of GNSS-based navigation in out-of-row scenarios ranged from 0.093 to 0.221 m, while the average heading deviation of in-row visual navigation was approximately 5.23° at a robot speed of 0.6 m/s. These results indicate that the proposed perception and navigation methods can maintain stable navigation performance within their respective applicable scenarios in trellis orchard environments. The experimental findings provide a practical and engineering-oriented basis for future research on automatic navigation mode switching and system-level integration of orchard transportation robots. Full article

With the intensification of global agricultural labor shortage and scaled development of facility agriculture, autonomous precision harvesting robots for unstructured greenhouse environments have become an urgent need. For cluster-picking crops such as tomatoes, visual servoing enables real-time closed-loop control of the end-effector pose, addressing challenges of random fruit distribution and variable stem orientations. However, existing methods struggle to balance constraint handling with real-time efficiency. This paper proposes an MPC-Guided Reinforcement Learning visual servoing framework, innovatively combining the planning capability of optimal control with the adaptive learning ability and real-time inference advantages of reinforcement learning. The approach adopts a teacher–student paradigm: expert trajectories from the MPC controller warm-start the reinforcement learning policy through behavior cloning, followed by PPO-based fine-tuning with adaptive gain regulation and stagnation-enhanced exploration mechanisms. Simulation experiments demonstrate a 95% success rate with average positioning and orientation errors of 13.6 mm and 0.009 rad respectively. Compared to MPC baseline, task steps are reduced by 53.4%; compared to Standard PPO, success rate improves by 6%. Greenhouse field validation achieves 85.3% picking success rate and 5.63 s per fruit operation time, confirming the framework’s excellent balance among control precision, robustness, and efficiency for high-precision robotic harvesting in unstructured agricultural environments. Full article

The tourism industry faces increasing pressure for agile, personalized services, yet travel agencies struggle with fragmented knowledge scattered across isolated systems and legacy formats. While Large Language Models (LLMs) are widely applied in customer-facing roles, their potential to enhance internal operational efficiency remains largely underexplored. This study presents the design and evaluation of an intelligent assistant specifically for travel agency operations, built upon a Retrieval-Augmented Generation (RAG) architecture using Gemini 2.0 Flash. The system integrates heterogeneous data sources, including structured product catalogs and unstructured documentation processed via Optical Character Recognition (OCR), into a unified interface comprising work assistance, interactive training, and evaluation modules. Results demonstrate information retrieval times not greater than 45 s, ensuring its daily usability, while maintaining 95% accuracy. Furthermore, the system democratizes tacit senior expertise and accelerates new employee onboarding. This research validates RAG architectures as a powerful solution to knowledge fragmentation, shifting the strategic AI focus from customer automation to employee empowerment and operational optimization. Full article

Non-Intrusive Load Monitoring (NILM) represents a powerful approach for energy disaggregation, which enables detailed insights into energy consumption patterns without requiring extensive sensor deployment. While significant advances have been achieved in residential NILM applications, commercial and industrial buildings remain largely underexplored despite their substantial contribution to global energy consumption. This study addresses this gap by developing and evaluating multiple artificial intelligence approaches for energy disaggregation across residential, commercial, and industrial buildings under a unified experimental protocol. We implement and compare several AI-based models, including Vision Transformer (ViT), Variational Autoencoder (VAE), Random Forest (RF), and custom architectures inspired by TimeGPT and Prophet, alongside traditional baseline methods. The proposed framework is validated using three benchmark datasets representing residential (AMPds), commercial (COmBED), and industrial (IMDELD) environments. Experimental results demonstrate that architecture–load interactions, rather than model complexity alone, are the primary determinants of disaggregation accuracy: the ViT-small configuration achieves superior performance for complex industrial loads with R² values exceeding 0.94, Random Forest proves most effective for finite-state commercial HVAC systems with R² up to 0.97, and the Prophet-inspired model excels in capturing seasonal patterns in residential appliances. These findings provide evidence-based guidelines for selecting appropriate AI models based on load characteristics, signal-to-noise ratio, and building type, contributing to the practical deployment of NILM in heterogeneous building environments. Full article

Background: Artificial Intelligence (AI) and Generative AI (GenAI) are increasingly integrated into educational and professional settings, offering personalized learning, productivity gains, and enhanced engagement. However, excessive reliance may compromise critical thinking, autonomous problem-solving, and emotional regulation among youth (i.e., adolescents and young adults) and early-career professionals. Aim: This review examines the cognitive, educational, and behavioral impacts of AI and GenAI use in youth, highlighting implications for their responsible integration in learning and professional development. Methods: A narrative review was conducted, synthesizing empirical studies, psychometric instruments, and international policy frameworks addressing AI engagement. Emphasis was placed on cognitive, behavioral, educational, and ethical dimensions across youth and early-career professionals. Results: AI enhances learning efficiency, creativity, and professional decision-making but may also foster cognitive offloading, dependency, and addiction-like behaviors. Instruments such as the Conversational AI Dependence Scale (CAIDS) and the Problematic ChatGPT Use Scale (PCGUS) help identify maladaptive patterns. Effective strategies include structured pedagogy, human oversight, reflective practice, AI literacy, and ethical guidance. Paradoxically, higher AI competence and trust may increase reliance, underscoring the need for guided and balanced engagement. Conclusions: Responsible AI integration requires multidimensional approaches combining instructional scaffolding, metacognitive strategies, supervision, and governance to preserve autonomy, professional judgment, and cognitive development in youth. Full article

The use of virtual environments to collect the experience required by deep reinforcement learning models is accelerating the deployment of these algorithms in industrial environments. However, once the experience-gathering problem is solved, it is necessary to address how to efficiently transfer the knowledge from the virtual scenario to reality. This paper focuses on examining Progressive Neural Networks (PNNs) as a promising transfer learning technique. The analyses carried out range from studying the capabilities and limits of the layers responsible for learning the state representation from a pixel space, which could arguably be the convolutional blocks, to the forgetting agents suffer when learning a new task. Introducing controlled visual changes in the environment scene can lead to a performance degradation of 50.3% in the worst-case scenario. These visual discrepancies significantly impact the agent’s learning time and accuracy when using a PNN architecture. Regarding the PNN forgetting assessment, partial forgetting occurs in two of the three environments analyzed, those where the agent masters its new task. This could be due to a balance between the relevance of the new features learned and the ones inherited from the teacher agent. Full article

This paper presents a comparative analysis of advanced AI-based techniques for human face inpainting using semantic masks that fully occlude targeted facial components. The primary objective is to evaluate the ability of image inpainting methods to accurately restore semantically meaningful facial features. Our results show that existing inpainting models face significant challenges when semantic masks completely obscure the underlying facial structures. In contrast to random masks, which leave partial visual cues, semantic masks remove all structural information, making reconstruction substantially more difficult. We assess the performance of generative adversarial networks (GANs), transformer-based models, and diffusion models in restoring fully occluded facial components. To address these challenges, we explore three retraining strategies: using semantic masks, using random masks, and a hybrid approach combining both. While the hybrid strategy leverages the complementary strengths of each mask type and improves contextual understanding, fully accurate reconstruction remains challenging. These findings demonstrate that inpainting under fully occluding semantic masks is a critical yet underexplored area, offering opportunities for developing new AI architectures and strategies for advanced facial reconstruction. Full article

Cross-view geo-localization tries to find the matching place in large satellite or aerial pictures from photos taken at ground level, which is useful for applications like self-driving cars, flying drones, and adding virtual objects to real city scenes. However, the traditional deep learning hybrid CNN-Transformer architecture and complex geometric submodules result in a large computational overhead, making it difficult to apply in real-time on resource-constrained devices. To make it light, fast, and accurate, this paper suggests an effective way to make a state-space model for cross-view geo-localization tasks. The model replaces the traditional self-attention structure with a state-space vision backbone, lowering the sequence modeling complexity from quadratic to linear and greatly accelerating the inference process; it devises a channel-group aggregation strategy without any learnable parameters, producing a comprehensive yet lightweight representation, and introduces a dynamic difficulty-aware loss function that assigns varying weights to all negative samples within a batch according to their similarities, greatly improving the efficiency of hard-negative sample mining and the quality of convergence. The results on the authoritative public datasets CVUSA and CVACT indicate that our method has high accuracy and low computational complexity, providing a feasible approach for the lightweight design of more powerful cross-view geolocation models in the future. Full article

This paper introduces Historical State Reconstruction, a novel framework for low-latency financial decision-making using Large Language Models. While agentic systems have demonstrated potential in synthesizing complex financial narratives, they typically rely on Retrieval-Augmented Generation or memory-based architectures. These paradigms introduce significant latency and risk look-ahead bias during real-time inference, rendering them unsuitable for high-frequency trading environments where milliseconds determine profitability. This proposed framework resolves this bottleneck by decoupling the heavy computational cost of context acquisition from the latency-sensitive critical path of decision-making. We propose a system that proactively compiles unstructured regulatory filings (10-K, 10-Q, 8-K) into a structured, bitemporal database. By pre-computing complex state facets, such as financial health ratios, governance structures, and insider trading signals offline, the system allows trading agents to “time travel” to a reconstructed state at any historical moment t with $O(1)$ snapshot retrieval plus $O(k)$ delta application complexity. We implement this approach on the top 50 companies in the S&P 500 ranked by market capitalization, processing over 12,000 filings to demonstrate a pipeline that transforms high-dimensional financial narratives into compact, prompt-ready context. Our evaluation shows that the system reduces context retrieval latency by over 97% compared to traditional baselines while achieving a 300:1 compression ratio for financial health data. Furthermore, the bitemporal architecture guarantees strict temporal integrity, eliminating the risk of data leakage in backtesting and satisfying the reproducibility requirements of regulatory frameworks like SR 11-7. Full article

Background: Deepfakes pose a growing threat to the integrity of visual media, motivating detectors that remain reliable as forgeries become increasingly realistic. Methods: We propose a deepfake detection framework built on CLIP-derived SigLIP-2 vision transformers and a multi-task design that jointly performs (i) classification and (ii) manipulated-region localization when pixel-level supervision is available. We evaluated the approach on three public benchmarks of increasing complexity—HiDF, SID_Set (SIDA), and CiFake—using each dataset’s official partitions where provided (SID_Set uses the predefined train/validation split) and a standardized preprocessing and training pipeline across experiments. Results: On HiDF, our model achieved strong performance on both video and image tracks (AUC up to 0.931 on video and 0.968 on images), yielding large gains relative to previously reported HiDF baselines under their published settings. On SID_Set, the model achieved 99.1% three-class accuracy (real/synthetic/tampered) and produced accurate localization masks for many tampered regions, while we explicitly documented the split protocol and leakage checks to support the validity of the evaluation. On CiFake, the model exceeded 95% accuracy and attained an AUC of 0.986. Conclusions: Overall, the results indicate that SigLIP-2 representations combined with multi-task training can deliver high detection accuracy and interpretable localization on challenging, realistic forgeries, while highlighting the importance of clearly stated evaluation protocols for fair comparison. Full article